The Hybrid Technique of Intrusion Cyber Attacks with Machine Learning Algorithms in IoT

Today, the Internet of Things (IOT) as a growing global network is susceptible to various attacks, and one of the dangerous attacks of the network layer is cyber attacks. Maintaining security against various cyber attacks within the network is considered as one of the important challenges of IoT. Intrusion detection system (IDS) is one of the main and effective defense methods to deal with attacks in IOT and plays an important role to identify and prevent cyber attacks in IoT networks. Different attacks have their own behavior, and attack detection using the combined method achieves a suitable performance in detecting new types of attacks. In this paper, a new hybrid method for attack and anomaly detection based on machine learning algorithms (Random Forest (RF), Perceptron Neural Network (MLP), Gradient Boosted Decision Trees (GBT) and K-Nearest Neighbor (K-NN)) is presented. It is proposed in the Internet of Things. Unlike the existing works that focus on single classifiers, this paper uses ensemble boosting and bagging algorithms to enhance the performance of intrusion detection system (IDS). The learning process and experiments have been performed on UNSW_NB15 and NLS_KDD datasets. The results show that random forest algorithms with accuracy (0.973 and 0.95) and Bagging with accuracy (0.998 and 0.997) effectively detect cyber attacks, and the ensemble algorithm of Bagging in terms of accuracy, recognition accuracy, recall and score. The F1 is better than comparable models.